Hi Georgi,

first of all, thank you very much for your response. Always glad to read from someone who seems to be more informed than me. :)

1) True that, nobody forced us to, and yes, I do indeed know that there are things like confidence intervals, thank you. But for reasons I can't exactly recall we just didn't do that. I think it was a mix of legacy setup + trying to make things easy for stakeholders.

2) Also true, but not my point. What I write is not about explaining the underlying concepts to stakeholders, independent of complexity, but interpretability of the results.

3) "Magical" in the sense of: it's just a convention that has developed during the 20th century in very different statistical contexts, isn't it? 95% is two standard deviations from the mean in a normal distribution, yes, but how often do you have that anyway? And, as I write, whenever I work together with economists, they use 90%, which simply is the convention in their field ...

4) If you read my article closely, I'm clearly not trying to fit p-values in a decision-making framework with Bayesian probabilities.

5) So, what exactly is wrong with the statements I cited from the Optimize resource hub? Can 1-p be interpreted as the probability that there is an uplift? And are p values intuitively understandable? Would be glad if you could elaborate.

Again, thanks for the comment and I wish you a great rest of the day.

Best,

Max

P.S.: I'm really, really sorry I made your fingers ache.